Improving methodology of quantifier comprehension experiments
Jakub Szymanik, Marcin Zajenkowski

Abstract:
Our investigation enriches and explains some data obtained by McMillan
and Troiani. We have shown that the computational model correctly
predicts that quantifiers computable by finite-automata are easier to
understand than quantifiers recognized by push-down automata. It
improves results of McMillan, which compared only first-order
quantifiers with higher-order quantifiers, putting in one group
quantifiers recognized by finite-automata as well as those recognized
by push-down automata. Moreover, we have assessed differences between
logical, parity, and numerical quantifiers.

Additionally, decreased reaction time in the case of proportional
quantifiers over ordered universes supports findings of McMillan, who
attributed the hardness of these quantifiers to the necessity of using
working memory.